RAPID COMMUNICATION www.jasn.org

A Single-Cell Transcriptome Atlas of the Mouse

Nikos Karaiskos,1 Mahdieh Rahmatollahi,2 Anastasiya Boltengagen,1 Haiyue Liu,1 Martin Hoehne ,2 Markus Rinschen,2,3 Bernhard Schermer,2,4,5 Thomas Benzing,2,4,5 Nikolaus Rajewsky,1 Christine Kocks ,1 Martin Kann,2 and Roman-Ulrich Müller 2,4,5

Due to the number of contributing authors, the affiliations are listed at the end of this article.

ABSTRACT Background Three different cell types constitute the glomerular filter: mesangial depending on cell location relative to the cells, endothelial cells, and . However, to what extent cellular heteroge- glomerular vascular pole.3 Because BP ad- neity exists within healthy glomerular cell populations remains unknown. aptation and mechanoadaptation of glo- merular cells are key determinants of Methods We used nanodroplet-based highly parallel transcriptional profiling to function and dysregulated in , characterize the cellular content of purified wild-type mouse glomeruli. we tested whether glomerular cell type sub- Results Unsupervised clustering of nearly 13,000 single-cell transcriptomes identi- sets can be identified by single-cell RNA fied the three known glomerular cell types. We provide a comprehensive online sequencing in wild-type glomeruli. This atlas of expression in glomerular cells that can be queried and visualized using technique allows for high-throughput tran- an interactive and freely available database. Novel marker for all glomerular scriptome profiling of individual cells and is cell types were identified and supported by immunohistochemistry images particularly suitable for identifying novel obtained from the Human Atlas. Subclustering of endothelial cells celltypesaswellassubsetsandnovelmarker revealed a subset of that expressed marker genes related to endo- genes of known cell populations.4–6 thelial proliferation. By comparison, the population appeared more ho- mogeneous but contained three smaller, previously unknown subpopulations. METHODS Conclusions Our study comprehensively characterized in individ- ual glomerular cells and sets the stage for the dissection of glomerular function at Glomerular isolation and preparation of the single-cell level in health and disease. single-cell suspensions were carried out J Am Soc Nephrol 29: 2060–2068, 2018. doi: https://doi.org/10.1681/ASN.2018030238

Significance Statement

The glomerular filtration unit consists of three Glomeruli are the key functional units of Received March 6, 2018. Accepted April 28, 2018. fi tightly intertwined cell types: the endothelium, the kidney ltration apparatus. Within N.K.,M.Rahmatollahi,M.K.,andR.-U.M.contrib- the mesangium, and podocytes. Despite es- each glomerulus, a tuft is struc- uted equally to this work. tablished physiologic cues acting on these cells within a single glomerulus, cellular heteroge- turally maintained by mesangial cells and Published online ahead of print. Publication date neity in the healthy glomerulus remains poorly provides a three-layered filtration barrier available at www.jasn.org. characterized. To address this problem, the consisting of endothelial cells, the glo- Correspondence: Dr. Christine Kocks, Systems Bi- authors performed large-scale transcriptional merular , and po- ology of Gene Regulatory Elements, Berlin Institute profiling of 13,000 mouse glomerular cells. docytes.1,2 Although these three cell for Medical Systems Biology, Max Delbrück Center They provide a comprehensive atlas of gene types within the glomerular tuft have for Molecular Medicine in the Helmholtz Associa- expression for the known glomerular cell types tion, Robert-Roessle-Str. 10, 13125 Berlin, Ger- and describe potential subpopulations for en- long been established, it is as yet un- many, or Dr. Roman-Ulrich Müller, Department II of dotheliumandpodocytes.Aninteractive,freely Internal Medicine and Center for Molecular Medi- known whether individual cells within accessible web tool allows for querying and the glomerulus respond to cues to which cine Cologne, University of Cologne, Kerpener Strasse 62, 50937 Cologne, Germany. Email: visualizing the data. The study highlights the they are physiologically exposed. Such [email protected] or roman-ulrich. power of single-cell RNA sequencing to study cues include changing pressure gradi- [email protected] gene expression in the kidney and sets the ents along the and mechanical stage for future investigations of glomerular Copyright © 2018 by the American Society of dysfunction in disease. strain on mesangial cells, which may differ Nephrology

2060 ISSN : 1046-6673/2908-2060 JAmSocNephrol29: 2060–2068, 2018 www.jasn.org RAPID COMMUNICATION as described7 on 8-week-old wild-type concentration used).4,8,12 To obtain by sequencing method, pairwise corre- CD1 male mice. Flow-sorted cells were high-quality single-cell data, we used a lations by cell type supported our cell dehydrated in methanol,8 stored and previously developed algorithm to score type assignments. shipped at 270°C,andrehydratedfor cell type–specific marker genes12 (Sup- We continued by characterizing glo- highly parallel single-cell transcriptome plemental Table 2) and removed 1768 merularcelltypesinmoredetailandaimed profiling by Drop-seq.4,8 This method probable doublets. The final dataset con- to identify novel cell-specificmarkersby predominantly detects 39 ends of polya- tained 12,954 cells, with a median of 630 assessing highly variable genes between denylated mRNA as well as long noncod- genes and 950 unique molecular identi- clusters.4 Established cell-specificmarker ing RNA molecules. Single-cell data were fiers per cell at a sequencing depth of genes for endothelium, mesangium, and processed, and genes were quantified approximately 9400 aligned reads per podocytes as well as genes described as with Drop-seq tools v. 1.124 and further cell (Supplemental Figure 1B, Supple- relevant to the respective cell type in the analyzed with “dropbead”8 and Seurat.5 mental Table 1). literature (Supplemental Methods has de- Marker gene identification was carried As shown in Figure 1B, unsupervised tails) were comprehensively reproduced out with Seurat function “FindAllMark- clustering of the remaining 12,954 single as specifically expressed, validating our ers”5 and visual inspection of violin plots cells identified five major cell clusters. unsupervised clustering (Figure 2A, Sup- as well as images from the Human Pro- On the basis of marker genes, three of plemental Figure 2).7,14,15 Importantly, tein Atlas.9 Immunofluorescence stain- these clusters corresponded to known expression of several previously reported ing was carried out on glomeruli of glomerular cell populations: podocytes key podocyte genes did not seem to be Nphs2-Cre/mTmG reporter mice10 using (80%), mesangial cells (2%), and endo- exclusive to podocytes, a finding bear- affinity-purified rabbit . Images thelial cells (12%). The other two clus- ing important implications for future were obtained using confocal microscopy. ters corresponded to tubular cells (6%) studies on the function of such genes Animal experiments were approved by the and a small group of immune cells in kidneys. Examples for such genes in- Landesamt für Natur, Umwelt und Verbrau- (0.2%). Glomerular cell type clusters clude Podxl (for which previously de- cherschutz Nordrhein-Westfalen (LANUV showed specificexpressionofestab- scribed endothelial expression was NRW, AZ 2013.A 375). Statistical methods lished marker genes (Figure 1, C and D, confirmed16), Actn4,andItgb1 (Figure were used as indicated. Supplemental Table 3), and all replicates 2, Supplemental Figure 2). Conse- Raw and processed datasets are avail- contributed to the observed cell clusters quently, we aimed to identify novel able from the Gene Expression Omnibus (Supplemental Figure 1C). Hierarchical cell-specificmarkersforallthreeglo- repository (GSE111107). The interactive clustering of aggregated reads from all merular cell types (Figure 2B). A large online database is available at https:// replicates and cell types indicated high proportion of these markers was cor- shiny.mdc-berlin.de/mgsca/. correlations according to cell type and roborated on the protein level by im- independent of the replicate (Supple- munostaining images obtained from mental Figure 1D). To control for effects the Human Protein Atlas (Figure 2B).9 RESULTS of single-cell preparation, we compared Novel markers represent a wide variety single-cell RNAseq data with bulk of molecular functions, including the Figure 1A shows the study design. We polyA-RNAseq libraries prepared from factor Meis2 identified as isolated glomeruli by magnetic bead per- glomeruli before and after dissociation specific to endothelial cells and disease fusion followed by magnetic separation into single cells (bulk1 and bulk2, re- genes, such as Pde3a (the gene mutated and rigorous washing (Supplemental spectively) (Figure 1A). Although the in autosomal dominant hypertension Figure 1A),11 generated single-cell sus- single-cell data showed good correla- with brachydactyly, which was identi- pensions by enzymatic digestion, and tions with both bulk mRNAseq datasets fied as specific to mesangial cells), as performed highly parallel single-cell (Supplemental Figure 1E), it became well as the E3-ubiquitin-ligase Wsb2,a RNA sequencing using the Drop-seq apparent that single-cell dissociation af- novel podocyte marker. Taken together, method.4 A total of 14,722 cells express- fected cell type abundance (Supplemen- we provide a detailed and comprehen- ing .250 genes and 350 transcripts (de- tal Figure 1, F and G), explaining an sive characterization of glomerular cell fined as unique molecular identifiers) over-representation of podocytes - types at the transcriptome level, in- were obtained from four independent tive to endothelial and mesangial cells cluding established and novel markers. biologic replicates (eight mice pooled (Figure 1B). (Tubules were not affected.) Thus far, glomerular gene expression per replicate). Median numbers of genes We also compared aggregated reads from has been examined almost exclusively in and transcripts detected were similar our cell type–specificclusterswithpub- cell populations rather than single cells. (Supplemental Figure 1B, Supplemental lished mRNAseq datasets obtained Therefore, it has remained unclear Table 1). Drop-seq works on the basis of from sorted cell populations on the ba- whether cell type heterogeneity exists Poisson-distributed limiting dilution, sis of glomerular cell lineage tracing ex- within the three glomerular cell types. and thus, it generates cell doublets (esti- periments7,13 (Supplemental Figure To approach this longstanding question, mated at approximately 10% at the cell 1H). Although samples correlated best we focused on subclustering the two larger

J Am Soc Nephrol 29: 2060–2068, 2018 A Glomerular Single-Cell Atlas 2061 RAPID COMMUNICATION www.jasn.org

Figure 1. Single-cell RNA-sequencing identifies the relevant cell populations in purified glomeruli. (A) Study design and workflow. (B) The plot shows a two-dimensional representation (tSNE: t-Distributed Stochastic Neighbor Embedding) of global relationships among ap- proximately 13,000 single cells expressing .350 transcripts (unique molecular identifiers). Putative cell doublets were removed by scoring cell type–exclusive markers (Supplemental Material). Five clusters became apparent that correspond to known cell types present in glomeruli (12% endothelium [n=1556], 2% mesangium [n=216], and 80% podocytes [n=10,325]) or contaminating cells from kidney tissue (6% tubules [n=828] and 0.2% immune cells [n=29]). Regarding parietal cells, none of the published marker genes were detected in any of the clusters. Cell types were identified by assessing the top most variable genes in each cluster (Supplemental Table 2). (C) Distribution and relative expression of established marker genes (violin plots) for endothelium, mesangium, podocytes, and contaminating tubular and immune cells.(D) Expression of marker genes colored on the basis of normalized expression levels (gray, low; red, high).

2062 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2060–2068, 2018 www.jasn.org RAPID COMMUNICATION

Figure 2. Single-cell transcriptomics reveal novel molecular markers specific to glomerular cell types. (A and B) Distribution and relative expression of individual highly variable genes (violin plots) in endothelium, mesangium, and podocytes. (A) Established markers (bold) and markers identified as relevant to the cell type in the literature (italics). (B) New marker genes identified in this study. (Left panel) Distribution and relative expression (violin plots). (Right panel) Immunohistochemistry images from the Human Protein Atlas (HPA) confirm that marker are expressed in human glomeruli in a histologic pattern as predicted from single-cell transcriptional analysis in mouse glomeruli. Image areas shown (5003500 pixels =200 mm2) correspond to glomeruli taken from larger HPA images.

J Am Soc Nephrol 29: 2060–2068, 2018 A Glomerular Single-Cell Atlas 2063 RAPID COMMUNICATION www.jasn.org clusters containing most cells—podocytes simply reflects localization in other parts responsive in podocytes,27 and it is hy- and endothelial cells. The latter showed of the kidney requires further investiga- pothesized to play a role in the develop- five distinct subclusters (Figure 3A, Sup- tion. ment of .28,29 plemental Table 3). Subcluster 4 was iden- Subcluster analysis in podocytes Lars2—a mitochondrial Leucyl transfer tified as residual cell doublets due to yielded seven subpopulations defined RNA synthetase—is mutated in Perrault expression of high levels of podocyte- by more subtle gene expression differ- syndrome.30 Although Perrault syndrome specificmarkers(Figure3B),anditwas ences (Supplemental Figure 4A, upper is primarily a neurologic disorder, muta- excluded from further analyses. The re- panel). In this context, cluster 4 showed tions of mitochondrial Leu-transfer RNA maining four subclusters showed equal an extensive stress response gene expres- are the basis of both Mitochondrial enceph- representation from all replicates (Sup- sion signature (Supplemental Figure 4A, alopathy, lactic acidosis, and stroke-like ep- plemental Figure 3, A and B). Expres- lower panel). Tissue dissociation–in- isodes syndrome (MELAS) syndrome and sion of key genes distinguished these duced changes in gene expression can hereditary FSGS,31,32 again pointing subclusters as illustrated in Figure 3, B explain this observation.26 Accordingly, toward a role in podocytes. andCandSupplementalFigure3C.In- we detected an increased expression of terestingly, subcluster 3 was defined by stress response genes in bulk mRNAseq marker genes implicated in key en- libraries obtained from dissociated glo- DISCUSSION dothelial molecular responses to phys- merular cells compared with whole glo- iologic and pathologic cues, such as meruli (Supplemental Figure 4, B and Our study also highlights a number of endothelium to crosstalk C). Reclustering of the podocytes after important caveats. First, we observed (Jag1),17–19 regulation of angiogenesis regressing out stress response genes an effect of the single-cell dissociation (Fbln5),20 endothelial activation identified six subclusters (Figure 4A, procedure. A comparison of our single- (Cxcl1),21 and response to complement Supplemental Table 3) with equal repre- cell data with bulk transcriptomes re- activation (Cldn5).22 Ehd3,amarker sentation of all biologic replicates (Sup- vealed an apparent over-representation suggested to be specific to glomerular plemental Figure 5, A and B). Three of podocytes relative to endothelial and endothelium,23,24 was unevenly ex- small subclusters (3–5) were identified mesangial cells as well as a stress re- pressed, with enrichment in subclus- robustly and independent of the stress sponse signature in one of the podocyte ters0and2andlowerexpressionin response signature (Figure 4A). Marker subclusters. As shown above, the latter subclusters 1 and 3; this raised the pos- gene analysis identified only a handful of kind of artifact can be corrected compu- sibility that a fraction of Ehd3-negative genes, including Cald1 and Lars2,aswell tationally. Second, although the vast cells in the endothelial pool originates as transcripts coding for mouse-specific majority of cells sequenced were clearly from other parts of the kidney. To ob- microRNAs (Gm10801 and Gm10800) glomerular, arguing for high purity of tain better functional understanding of and noncoding RNAs (Gm15564 and isolated glomeruli, an extraglomerular the endothelial subclusters, we per- Gm23935)(Figure4,BandC,Supple- origin for a fraction of endothelial cells formed pathway and gene set overdis- mental Figure 5C), whereas the re- ispossible.Third,weexaminedmale persion analysis.25 Four gene sets were maining three larger clusters did not mice of one strain at an age when glo- identified that characterized the sub- yield specificmarkers.Giventhesmall meruli are still enlarging. Thus, the glo- clusters to varying extent with terms number of coding transcripts among merular subpopulations observed may relating to “cell adhesion,”“cell matu- subcluster markers, pathway and gene not necessarily be stable in mice of dif- ration,”“stress response,” and “cell pro- set overdispersion analysis did not ferent ages, sexes, or strains. liferation” (Figure 3D, Supplemental yield significant results (Supplemental In summary, our study comprehen- Figure 3E, Supplemental Table 4), sug- Figure 6). sively characterizes gene expression in gesting that the endothelial subclusters To corroborate podocyte subcluster individual glomerular cells. We identi- might represent different states of en- markers on the protein level, immunoflu- fied novel marker genes for all glomeru- dothelial cell biology between homeo- orescence staining was carried out for lar cell types and found evidence for stasis and activation. Interestingly, Cald1 and Lars2 on glomeruli obtained transcriptional heterogeneity among en- staining in the Human Pro- from reporter mice, in which podocytes dothelial cells and podocytes. Earlier tein Atlas for some proteins encoded are marked by green fluorescent pro- publications using single-cell RNA tran- by the subcluster determining genes, tein.10 Colocalization of Cald1 and Lars2 scriptomics on glomerular cells were such as Fbln2, Hspa1b, S100a4,andThbd with green fluorescent protein occurred limited by focusing exclusivelyon a single (Figure 3B), was consistent with a non- only in a subset of podocytes (Figure 4D, cell type and sequencing small numbers uniform expression pattern of these genes Supplemental Figure 5D), suggesting that of cells (20 podocytes and 14 mesangial in human glomerular endothelial cells heterogeneity among podocytes in cells, respectively).33,34 In contrast, our (Supplemental Figure 3D). Whether this healthy glomeruli might exist. Cald1 is a approach has exploited the potential of state depends on individual cell localiza- and binding protein that highly parallel single-cell profiling for tion within a healthy capillary tuft or has been shown to be glucocorticoid profiling a large number of cells in an

2064 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2060–2068, 2018 www.jasn.org RAPID COMMUNICATION

Figure 3. Subclustering reveals the presence of endothelial subpopulations. (A) Two-dimensional representation of a subclustering analysis of endothelial cells. Five subclusters (0–4) became apparent. (B) Distribution and relative expression of individual highly variable genes (violin plots) in the different clusters. Cluster 4 corresponds to residual cell doublets as indicated by the expression of podocyte- specific markers (Nphs2 and Cdkn1c). Doublets were excluded from further analysis. (C) Expression of markers colored on the basis of normalized expression levels. Upper panels correspond to the subcluster tSNE (t-Distributed Stochastic Neighbor Embedding) plot as shown in A, and lower panels correspond to the tSNE plot of the whole dataset as shown in Figure 1B. (D) Pathway and gene set over- dispersion analysis.33 The heat map indicates four endothelial subclusters (0, red; 1, green; 2, blue; 3, violet) that show distinct, over- represented gene activation patterns (Supplemental Figure 3E). Corresponding gene clusters are listed in Supplemental Table 4.

J Am Soc Nephrol 29: 2060–2068, 2018 A Glomerular Single-Cell Atlas 2065 RAPID COMMUNICATION www.jasn.org

paves the way for future investigations addressing the response of individual glomerular cells to disease states—as illustrated by a pilot study on kidney biopsy specimens from patients with lupus .35 Theglomerulusisa key system, the understanding of which will greatly benefitfromim- proved single-cell RNA sequencing protocols.

ACKNOWLEDGMENTS

We thank Martyna Brütting for excellent technical support with immunofluorescence staining and immunohistochemistry. Chris- tian Jüngst (Cologne Excellence Cluster on Cellular Stress Responses in Aging-Associ- ated Diseases Imaging Core Facility) pro- vided excellent support regarding confocal microscopy, and Gunter Rappl (CMMC Fluorescence-activated cell sorting Facility) helped with cell sorting. This work was supported by German Re- search Foundation grants RA838/5-1 (to N.R.), KA3217/4-1 (to M.K.), KFO329 (to M.K. and R.-U.M.), and MU3629/2-1 (to R.-U.M.); Berlin Institute of Health grant CRG2aTP7 (to N.R.); Deutsches Zentrum fuer Herz-Kreislauffor- schung e.V. grant BER1.2VD (to N.R.); the Helmholtz Association through Helmholtz Ex- cellence Network for NeuroCure grant HFG ExNet-0036-phase2-3 (to N.R.); and the Nachwuchsgruppen Nordrhein-Westfalen Pro- Figure 4. Subclustering reveals a limited heterogeneity of podocytes. (A) Two-dimensional gram of the Ministery of Science North Rhine representation of a subclustering analysis of podocytes (tSNE: t-Distributed Stochastic Westfalia (R.-U.M.). M. Rahmatollahi was Neighbor Embedding) after correction for tissue dissociation–induced stress response supported by the Graduate Program in Phar- gene expression (Supplemental Material). Six podocyte subclusters (0–5) became appar- macology and Experimental Therapeutics at the ent. (B) Distribution and relative expression of individual highly variable genes (violin plots) University of Cologne, which is financially and in subcluster 4. (C) Expression of markers in subcluster 4 (corresponding to B). Expression scientifically supported by Bayer. colored is on the basis of normalized expression levels (gray, low; red, high). (D) Laser- T.B., B.S., and N.R. conceived the study; scanning confocal microscopy of isolated glomeruli from kidneys of transgenic Nphs2- N.R., M.K., and R.-U.M. procured funding; Cre3mT/mG double-fluorescent reporter mice.35 Podocytes are genetically marked by N.K., M.Ra., C.K., M.K., and R.-U.M. designed Cre-dependent membrane-targeted green fluorescent protein [GFP] (green) fluorescence, thestudy;N.R.supervised,N.K.performedall whereas nonpodocyte cell types remain membrane-targeted Tomato (red) positive. (Row 1) Whole glomeruli; yellow squares in podocyte staining (green) indicate areas for mag- computational analyses and designed the online nifications as shown below. Cald1 antibody staining (upper square) and IgG control (lower database, and H.L. performed PAGODA; M.Ra. square). (Rows 2 and 3) Insets from whole glomeruli as indicated. Yellow arrowheads point carried out mouse experiments and microscopy; to GFP-positive podocytes that are Cald1 negative (row 2) or unstained by IgG control (row M.H. analyzed imaging data; C.K. supervised, 3). (Rows 4 and 5) Yellow arrowheads point to a GFP-positive, Cald1-positive podocyte. A.B.performedsingle-cellandbulkmRNAse- Magnified areas are 22322 mm2. Scale bars: 10 mm. quencing; N.K., M.Ra., M.H., M.Ri., C.K., M.K., and R.-U.M. analyzed and discussed data; N.K., unbiased way.33,34 As a resource, we pro- transcriptome that can be freely accessed C.K., M.H., and M.K. prepared the figures; vide an extensive single-cell sequencing and interrogated online (https://shiny. N.K., C.K., M.K., and R.-U.M. wrote the man- dataset of the mouse glomerular mdc-berlin.de/mgsca/). Our study thus uscript; all authors approved the final version.

2066 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2060–2068, 2018 www.jasn.org RAPID COMMUNICATION

DISCLOSURES fingerprinting of the podocyte reveals novel 26. van den Brink SC, Sage F, Vértesy Á, Spanjaard gene and protein regulatory networks. Kid- B, Peterson-Maduro J, Baron CS, et al.: Single- None. ney Int 83: 1052–1064, 2013 cell sequencing reveals dissociation-induced 15. Potter SS, Brunskill EW: Building an atlas of gene expression in tissue subpopulations. Nat gene expression driving kidney develop- Methods 14: 935–936, 2017 REFERENCES ment: Pushing the limits of resolution. Pe- 27. Jiang L, Hindmarch CCT, Rogers M, diatr Nephrol 29: 581–588, 2014 Campbell C, Waterfall C, Coghill J, et al.: 1. Brinkkoetter PT, Ising C, Benzing T: The role 16. Horrillo A, Porras G, Ayuso MS, González- RNA sequencing analysis of human podo- of the podocyte in filtration. Nat Rev Manchón C: Loss of endothelial barrier in- cytes reveals glucocorticoid regulated gene Nephrol 9: 328–336, 2013 tegrity in mice with conditional ablation of networks targeting non-immune pathways. 2. Scott RP, Quaggin SE: Review series: The cell podocalyxin (Podxl) in endothelial cells. Eur J Sci Rep 6: 35671, 2016 biology of renal filtration. J Cell Biol 209: Cell Biol 95: 265–276, 2016 28. Wang Z, Wang Z, Zhou Z, Ren Y: Crucial 199–210, 2015 17. Cao Z, Lis R, Ginsberg M, Chavez D, Shido K, genes associated with diabetic nephropathy 3. Neal CR, Muston PR, Njegovan D, Verrill R, Rabbany SY, et al.: Targeting of the pulmo- explored by microarray analysis. BMC Harper SJ, Deen WM, et al.: Glomerular fil- nary capillary vascular niche promotes lung Nephrol 17: 128, 2016 fi  ł tration into the subpodocyte space is highly alveolar repair and ameliorates brosis. Nat 29. Snit M, Nabrdalik K, D ugaszek M, –  restricted under physiological perfusion Med 22: 154 162, 2016 Gumprecht J, Trautsolt W, Górczynska- conditions. Am J Physiol Renal Physiol 293: 18. Liu H, Kennard S, Lilly B: NOTCH3 ex- Kosiorz S, et al.: Association of rs 3807337 F1787–F1798, 2007 pression is induced in mural cells through polymorphism of CALD1 gene with diabetic 4. Macosko EZ, Basu A, Satija R, Nemesh J, an autoregulatory loop that requires nephropathy occurrence in type 1 diabetes - Shekhar K, Goldman M, et al.: Highly parallel endothelial-expressed JAGGED1. Circ preliminary results of a family-based study. – – genome-wide expression profiling of indi- Res 104: 466 475, 2009 Endokrynol Pol 68: 13 17, 2017 vidual cells using nanoliter droplets. Cell 161: 19. Scheppke L, Murphy EA, Zarpellon A, 30. Pierce SB, Gersak K, Michaelson-Cohen R, 1202–1214, 2015 Hofmann JJ, Merkulova A, Shields DJ, et al.: Walsh T, Lee MK, Malach D, et al.: Muta- 5. Satija R, Farrell JA, Gennert D, Schier AF, Notch promotes vascular maturation by in- tions in LARS2, encoding mitochondrial Regev A: Spatial reconstruction of single-cell ducing integrin-mediated leucyl-tRNA synthetase, lead to premature gene expression data. Nat Biotechnol 33: cell adhesion to the endothelial basement ovarian failure and hearing loss in Perrault – 495–502, 2015 membrane. 119: 2149–2158, 2012 syndrome. Am J Hum Genet 92: 614 620, 6. Chen L, Lee JW, Chou C-L, Nair AV, Battistone 20. Chan W, Ismail H, Mayaki D, Sanchez V, 2013 MA, Păunescu TG, et al.: Transcriptomes of TiedemannK,DavisEC,etal.:-5regu- 31. Cheong HI, Chae JH, Kim JS, Park HW, Ha major renal collecting duct cell types in mouse lates Angiopoietin-1/Tie-2 signaling in IS, Hwang YS, et al.: Hereditary glomerul- identified by single-cell RNA-seq. Proc Natl endothelial cells. PLoS One 11: e0156994, 2016 opathy associated with a mitochondrial Acad Sci U S A 114: E9989–E9998, 2017 21. Mai J, Nanayakkara G, Lopez-Pastrana J, Li X, tRNA(Leu) gene mutation. Pediatr Nephrol 13: – 7. Kann M, Ettou S, Jung YL, Lenz MO, Taglienti Li Y-F, Wang X, et al.: -17A pro- 477 480, 1999 ME, Park PJ, et al.: Genome-wide analysis of motes aortic endothelial cell activation via 32. Jansen JJ, Maassen JA, van der Woude FJ, ’ wilms’ tumor 1-controlled gene expression in transcriptionally and post-translationally acti- Lemmink HA, van den Ouweland JM, t Hart podocytes reveals key regulatory mechanisms. J vating p38 Mitogen-activated Protein LM, et al.: Mutation in mitochondrial tRNA Am Soc Nephrol 26: 2097–2104, 2015 (MAPK) pathway. JBiolChem291: 4939–4954, (Leu(UUR)) gene associated with progressive 8. Alles J, Karaiskos N, Praktiknjo SD, Grosswendt 2016 kidney disease. JAmSocNephrol8: – S,WahleP,RuffaultP-L,etal.:Cellfixation and 22. Dalmasso AP, Goldish D, Benson BA, Tsai 1118 1124, 1997 preservation for droplet-based single-cell tran- AK, Wasiluk KR, Vercellotti GM: In- 33. Lu Y, Ye Y, Bao W, Yang Q, Wang J, Liu Z, fi scriptomics. BMC Biol 15: 44, 2017 terleukin-4 induces up-regulation of en- et al.: Genome-wide identi cation of genes 9. Uhlén M, Fagerberg L, Hallström BM, dothelial cell claudin-5 through activation essential for podocyte cytoskeletons based Lindskog C, Oksvold P, Mardinoglu A, et al.: of FoxO1: Role in protection from com- on single-cell RNA sequencing. Kidney Int – Proteomics. Tissue-based map of the human plement-mediated injury. J Biol Chem 289: 92: 1119 1129, 2017 proteome. Science 347: 1260419, 2015 838–847, 2014 34. Lu Y, Ye Y, Yang Q, Shi S: Single-cell RNA- 10. Muzumdar MD, Tasic B, Miyamichi K, Li L, 23. George M, Rainey MA, Naramura M, Foster sequence analysis of mouse glomerular Luo L: A global double-fluorescent Cre re- KW, Holzapfel MS, Willoughby LL, et al.: mesangial cells uncovers mesangial cell es- – porter mouse. Genesis 45: 593–605, 2007 Renal thrombotic microangiopathy in mice sential genes. Kidney Int 92: 504 513, 2017 11. Takemoto M, Asker N, Gerhardt H, Lundkvist A, with combined deletion of endocytic re- 35. Der E, Ranabothu S, Suryawanshi H, Akat KM, Johansson BR, Saito Y, et al.: A new method for cycling regulators EHD3 and EHD4. PLoS ClancyR,MorozovP,etal.:SinglecellRNAse- large scale isolation of kidney glomeruli from One 6: e17838, 2011 quencing to dissect the molecular heterogene- mice. Am J Pathol 161: 799–805, 2002 24. Patrakka J, Xiao Z, Nukui M, Takemoto M, He ity in . JCI Insight 2: 2, 2017 12. Karaiskos N, Wahle P, Alles J, Boltengagen L, Oddsson A, et al.: Expression and sub- A, Ayoub S, Kipar C, et al.: The Drosophila cellular distribution of novel glomerulus-as- embryo at single-cell transcriptome resolu- sociated proteins dendrin, ehd3, sh2d4a, “ tion. Science 358: 194–199, 2017 plekhh2, and 2310066E14Rik. JAmSoc See related perspective, Single-Cell Sequencing the Glomerulus, Unraveling the Molecular Pro- 13. Brunskill EW, Park J-S, Chung E, Chen F, Nephrol 18: 689–697, 2007 grams of Glomerular Filtration, One Cell at a Time,” Magella B, Potter SS: Single cell dissection of 25. Fan J, Salathia N, Liu R, Kaeser GE, Yung YC, on pages 2036–2038. early : Multilineage prim- Herman JL, et al.: Characterizing transcriptional – ing. Development 141: 3093 3101, 2014 heterogeneity through pathway and gene set This article contains supplemental material online at 14. Boerries M, Grahammer F, Eiselein S, Buck overdispersion analysis. Nat Methods 13: 241– http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ M,MeyerC,GoedelM,etal.:Molecular 244, 2016 ASN.2018030238/-/DCSupplemental.

J Am Soc Nephrol 29: 2060–2068, 2018 A Glomerular Single-Cell Atlas 2067 RAPID COMMUNICATION www.jasn.org

AFFILIATIONS

1Systems Biology of Gene Regulatory Elements, Berlin Institute for Medical Systems Biology, Max Delbrück Center for Molecular Medicine in the Helmholtz Association, Berlin, Germany; 2Department II of Internal Medicine and Center for Molecular Medicine Cologne, 4Cologne Excellence Cluster on Cellular Stress Responses in Aging-Associated Diseases, and 5Systems Biology of Ageing Cologne, University of Cologne, Cologne, Germany; and 3Scripps Center for Metabolomics and Mass Spectrometry, The Scripps Research Institute, San Diego, California

2068 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2060–2068, 2018 Supplemental Material (online)

Title: A Single-cell Transcriptome Atlas of the Mouse Glomerulus

Running title: Single-cell transcriptomics of glomerular cells

Authors: Nikos Karaiskos, Mahdieh Rahmatollahi, Anastasiya Boltengagen, Haiyue Liu, Martin Hoehne, Markus Rinschen, Bernhard Schermer, Thomas Benzing, Nikolaus Rajewsky, Christine Kocks, Martin Kann, Roman-Ulrich Müller.

Table of contents - Supplemental Tables and Figures...... page 1

Supplemental Methods ...... page 2

Supplemental Tables and Figures ...... page 6

Supplemental Figure and Table legends ...... page 17

References for Supplemental Material ...... page 20

Supplemental Tables and Figures

Supplemental Figure 1. (Related to Fig. 1) Quality of glomeruli single-cell data and correlations with bulk mRNAseq data

Supplemental Table 1. (Related to Suppl. Fig. 1B) Median number of genes, transcripts and reads per cell

Supplemental Table 2. (Related to Suppl. Fig. 1B) Marker genes for each of the five major cell types to identify putative cell doublets

Supplemental Table 3. (Related to Fig. 1B, Fig. 3A and Fig. 4A) Top variable genes per cluster

Supplemental Figure 2. (Related to Fig. 2) Additional markers of glomerular cells

Supplemental Figure 3. (Related to Fig. 3) Subclustering of glomerular endothelium

Supplemental Table 4. (Related to Fig. 3D and Suppl. Fig. 3E) Gene sets underlying the (GO) term analysis for endothelium subclusters

Supplemental Figure 4. (Related to Fig. 4) Stress response in podocytes

Supplemental Figure 5. (Related to Fig. 4) Subclustering of podocytes

Supplemental Figure 6. (Related to Fig. 4) Pathway and gene set overdispersion analysis for podocyte subclusters

1 Supplemental Methods

Glomerular isolation and single cell suspension Isolation of glomeruli and preparation of single-cell suspensions from 8 week old CD1 male mice was performed as previously described1. In brief, kidneys were dissected after cervical dislocation and renal arteries were perfused with Hank’s Balanced Salt Solution (HBSS) supplemented with magnetic Dynabeads (Thermofisher Sicentific). Next, kidneys were released from the capsule and minced in 3 ml tissue digestion solution containing 300 U/ml Collagenase 4 (Worthington), 300 U/ml Protease and 50 U/ml DNase I. Tissue dissociation was completed in a 37 ºC incubator with gentle shaking and trituration in 5-minute intervals to facilitate the dissociation process. The tissue soup was then passed through 100-um sieve on 50-ml falcon tubes to capture tubular and non-glomeruli structures by their size. The sieves were washed with ice-cold HBSS and the glomeruli were pelleted by spinning down at 1500x g at 4 ºC for 5 minutes. The pellet was resuspended in ice-cold HBSS in 1.5 ml eppendorf tubes and the glomeruli were concentrated using a magnet. The purity of glomerular isolates was routinely monitored by visual inspection on a Leica DFC450 microscope (Suppl. Figure 1A). For preparation of single cells, isolated mouse glomeruli were re-suspended in 1 ml of tissue digestion solution and transferred to a thermo- set to 37 °C at1400 rpm. Further mechanical dissociation was induced using pipetting and/or gentle shearing through a syringe in 5-minute intervals. After a total time of 40 minutes, beads were removed and the single cells were collected by centrifugation at 5000x g, 4 ºC for 10 minutes. Cells were re-suspended in ice-cold PBS and immediately FACS-sorted. 7-AAD (BD Pharmigen, 559925) was added to stain dead cells. FACS was carried out on a FACSAria III device gating for 7-AAD negative single cells. For fixation, single live cells were collected in ice-cold 80% methanol in PBS and transferred to Berlin on dry ice for the preparation of monodisperse droplets for single cell RNAseq.

Single-cell and bulk mRNA-sequencing

Single-cell sequencing: Drop-seq procedure, single-cell library generation and sequencing Monodisperse droplets of about 1 nl in size were prepared on a self-built Drop-seq set up following closely the instrument set up and library generation procedures invented by Macosko et al.2 , as also previously described by us for methanol-fixed cells3,4. Upon nanoliter droplet formation, individual cells are co-encapsulated with individual, uniquely barcoded beads, and become lysed completely. Released cellular polyadenylated RNA molecules then hybridize to polyd(T) primers that are attached to the uniquely barcoded beads. Nanoliter-droplets are collected and broken, and RNA molecules are reverse transcribed into complementary DNA (cDNA), amplified by PCR, size-selected for mRNA and sequenced in bulk. cDNA libraries had large average sizes of ~ 1.7 to 1.9 kb, indicating high quality RNA and cDNA molecules. Single-cell Drop-seq libraries at 1.8 pM (final insert size average ranging between 650 to 780 bp) were sequenced in paired-end mode on Illumina Nextseq 500 sequencers with 1% PhiX spike-in for run quality control using Illumina Nextseq500/550 High Output v2 kits (75 cycles). Read 1: 20 bp (bases 1-12 cell barcode, bases 13-20 UMI; Drop- seq custom primer 1 “Read1CustSeqB”), index read: 8 bp, read 2 (paired end): 64 bp).

Bulk mRNA-seq libraries and sequencing Isolated glomeruli or live, sorted, methanol-fixed single glomeruli cells were used for total RNA extraction with TRIzol reagent (Invitrogen). Stand-specific cDNA libraries were generated according to the Illumina TruSeq protocol (TruSeq Stranded mRNA LT Sample Prep Kit, Illumina) using 500 ng (glomeruli lysate) or 95 ng (sorted, fixed single cells from dissociated glomeri) of total RNA. 1.8 pM concentrated libraries were sequenced in single-read mode on Illumina Nextseq 500 sequencers using High Output v2 (150 cycles), single read 150 bp, index read: 6 bp.

2

Computational methods

Bulk mRNA-seq: alignment and gene quantification. Reads were aligned to the Mus musculus reference genome mm10 using STAR v.2.5.3a with default parameters5. Gene quantification was done with htseq-count by using the GENCODE M12 (Ensembl 87) annotation. We used the same workflow to align the reads and quantify genes in published datasets (Suppl. Fig. 1F-H).

Single cell RNA-seq: data processing, alignment and gene quantification. We chose read 1 to be 20bp long, which sequences the cell barcode at positions 1-12 and the UMI at positions 13-20, while avoiding reading into the poly(A) tail. The remaining 64 sequencing cycles were used for read 2. Sequencing quality was assessed by FastQC (v.0.11.2), while special attention was paid to the base qualities of read 1 to assure accurate cell and UMI calling. We used the Drop-seq tools v. 1.122 to trim poly(A) stretches and potential SMART adapter contaminants from read 2, to add the cell and molecular barcodes to the sequences and to filter out barcodes with low quality bases. The reads were then aligned to the reference genome as above. Typically, around 60%-75% of the reads were found to uniquely map to the genome; muti-mapping reads were discarded. The Drop-seq toolkit2 was further used to add gene annotation tags to the aligned reads and to identify and correct bead synthesis errors, in particular base missing cases in the cell barcode. Cell numbers were estimated by plotting the cumulative fraction of reads per cell against the cell barcodes and calculating the knee point. The DigitalExpression tool2 was used to obtain the digital gene expression matrix (DGE) for each sample.

Cell filtering and data normalization. For a first view of gene quantification and basic statistics we used the R package 'dropbead' (https://github.com/rajewsky-lab/dropbead)3. The individual DGEs corresponding to four replicate samples were pooled together, resulting in 15,135 cells. We discarded cells expressing less than 350 UMIs, ending up with 14,722 for further analysis. We normalized the UMI counts for every gene per cell by dividing its UMI count by the sum of total UMIs in that cell, and multiplying it by the number of UMIs that the deepest cell contained. Downstream analysis was performed in log space.

Doublet identification and removal. We used the marker genes in Table 1 and the following algorithm to identify and remove cell doublets4. First, for each of the marker genes, we computed its mean expression only from cells which express it. Second, for a given cell, we used these computed averaged values to scale the expression of any of the marker genes expressed in that cell, producing scaled scores. This has the effect of not favoring outlier, or very highly expressed genes. Next, we computed the mean score of every cell type for every cell. If a cell had two or more cell type scores larger than a cutoff, it was flagged as a doublet. As we expected around ~10% of doublets based on earlier cell mixing experiments with distinct species3,4, we took a threshold value of 0.08, identifying 1,754 doublets in our data. We removed those cells to continue our analysis.

Correlation of gene expression measurements. Correlations of gene expression levels between single-cell samples were computed by first sub-setting the DGEs of the two samples to the intersection of the genes captured in both libraries (typically more than ~15,000) and then computing the sum of gene counts across all cells in each library. Plotting of correlations is shown in log-space. For the correlation of our

3 single-cell libraries against mRNA-seq, we converted the gene counts of the latter one into RPKMs and used the mean value of all isoforms for a given gene.

Clustering, t-SNE representation and marker discovery. We used Seurat6 v. 2.2.0 to identify highly variable genes, perform principal component analysis, identify the most important principal components and use those for clustering and t- SNE representation. Around 1305 highly variable genes were identified and the first 15 principal components were used for the clustering and the tSNE representation. Cluster identification was done with the Louvain clustering algorithm. The endothelial and podocyte subclusters were found to be relatively robust to changes of the resolution parameter. We used Seurat’s function FindAllMarkers to find potential markers for each cluster in an unsupervised manner.

Detection of tissue dissociation-induced stress response A set of 140 genes, recently shown to respond to tissue dissociation stress, was taken from van den Brink et al. 20177 (Suppl. Table 5). In order to assess whether our single-cell dissociation induced a stress response, we compared our two bulk mRNA-seq samples ("bulk1" corresponding to purified glomeruli; "bulk2" corresponding to sorted, fixed single cells). The difference of normalized gene expression between the two samples was used to measure the stress response. We restricted the analysis to those genes having a normalized expression greater than 1 in both bulk samples, resulting in 132 out of 140 initial genes. We created a second set by randomly selecting an equal number of genes with the same requirement. A third set of genes was created by excluding the stress response genes from the highly variable genes and requiring again a normalized expression greater than 1 (975 genes in total). Normality of the three distributions was assessed with a Kolmogorov-Smirnov test (p<2.2x10-16, p=1.9x10-4, p<2.2x10-16). The differences in means between the three sets was found to be statistically significant, if tested with a t-test (p<2.2x10-16 for each comparison).

Subclustering and correction for tissue dissociation-induced stress response genes We subclustered the podocytes and the endothelial cells by extracting them from the main clustering and treating them as independent samples. The corresponding raw DGEs (with UMI counts) were used as input to Seurat. As the podocytes subclusters were defined by subtle differences in gene expression, we regressed out the stress response genes to alleviate any stress-related confounding factors. We found that the removal of those genes from the highly variable genes of the Seurat object was the most effective way to regress out the stress factor.

Functional annotation by gene ontology analysis We used pathway and gene set overdisperson analysis (PAGODA)8 to detect functional characteristics across cell populations in endothelial and podocyte subclusters. The color bars above PAGODA heatmaps (Fig. 3C, Suppl. Fig. 4H) were defined based on cell cluster colors shown in the tsne plots (Fig. 3A, 4A). Both predefined KEGG pathways and de novo gene clusters were tested. For significant over-dispersed de novo gene clusters, gene ontology analysis was performed using clusterProfiler9 with pAdjustMethod="BH", pvalueCutoff=0.01, and qvalue = 0.05.

Image analysis and microscopy

Literature and Human Protein Atlas review for identification of novel marker genes Cell cluster marker genes as identified by highly variable gene analysis between clusters were classified in three categories: ‘established marker’, ‘relevant as by published literature’, and ‘novel’. The search strategy was designed to identify a comprehensive list of novel cell-type specific markers for the three key glomerular cell types. Ideally, such novel markers would be corroborated by immunostaining data from the Human Protein Atlas (HPA)10.

4 In a first step, disease genes causing hereditary kidney disease upon mutation and genes widely used as markers for the respective cell population in immunohistochemistry and immunofluorescence staining were classified as ‘established marker’. Then, violin plots for marker expression in glomerular cells were visually inspected to identify marker genes showing exclusive expression in either podocytes, mesangial, or endothelial cells. Genes that either did not have corresponding data in HPA, showed ubiquitous expression in HPA, or were classified as ‘not detected’ or ‘low expression’ in glomeruli within HPA, respectively, were not followed up further. The remaining genes were searched for in the Pubmed database as current in November 2017 using “[(gene symbol) OR (gene name) OR (protein name)] AND endothel* AND (kidney OR glomerulus)”, “[(gene symbol) OR (gene name) OR (protein name)] AND podocyt* AND (kidney OR glomerulus)”, and “[(gene symbol) OR (gene name) OR (protein name)] AND mesangi* AND (kidney OR glomerulus)”, respectively, as search terms. Upon identification of at least one abstract relating to a function of the gene in question in glomeruli or kidneys in vivo, the respective gene was classified as ‘relevant as by published literature’. If no publications were identified in this manner Pubmed was searched again using “[(gene symbol) OR (gene name) OR (protein name)] AND (kidney OR glomerulus)” as search term. If no publications were found or the putative marker gene had thus far only been shown as relevant in vitro, the marker was classified as ‘novel’.

Immunofluorescence staining & image processing Immunofluorescent staining of isolated glomeruli was carried out on 8 week old male Nphs2- Cre x mTmG mice in a CD1 background11. Fixation was performed with 10% PFA followed by permeabilization with Triton-based PBS at room temperature for 30 minutes. Primary antibody incubation was completed overnight at 4 ºC with Cald1 antibody (Sigma, HPA008066), Lars2 antibody (Abcam, ab187983) or an isotype control (Santa Cruz, sc-2027). Confocal stacks were acquired using a Leica SP8 confocal microscope equipped with a 63x/1.4 oil immersion objective operated with the LAS X software version 3.1.5 (Leica Microsystems, Wetzlar, Germany). Image acquisition settings were comparable between all images. To prepare representative figures images were processed using ImageJ/Fiji (version 1.51) as follows: Maximum intensity projections of 5 slices spanning 1.49µm. Display settings (Min&Max adjustement) in the control channels (mTomato, mGFP, DAPI) were individually adjusted for best result, while the settings were kept identical for the channel of interest (Cald1, Lars2 and control). Montages were prepared using the ImageJ/Fiji Montage tool.

5 Supplemental Figure 1

A B

biological replicates

C D Tubules Podocytes Endothelium Mesangium

Podocytes

Tubules

Replicate

Immune cells

Mesangium Endothelium

E F Supplemental Figure 1

G H

bulk single-cell Sheet1 Supplemental Table 1. Median number of genes, transcripts (UMIs) and reads per cell

reads cells genes umis before replicate1 8077 4394 532 786 doublet replicate2 10091 4164 743 1156 removal replicate3 8982 2948 595 884 replicate4 10562 3216 649 986 mean 9428 3681 630 953 sum 14722 after replicate1 8294 3815 544 806 doublet replicate2 9744 3768 730 1127 removal replicate3 8963 2553 589 878 replicate4 10496 2818 648 983 mean 9374 3239 628 949 sum 12954

Page 1 Supplemental Table 2. (Related to Suppl. Fig. 1B). Marker genes for each of the five major cell types to identify putative cell doublets.

EndotheliumMesangium Podocytes Tubules Immune Emcn Acta2 Nphs2 Fxyd2 Cd74 Ehd3 Rgs5 Cdkn1c Pvalb H2-Aa Kdr Myl9 Tcf21 Atp1b1 Cd52 Flt1 Tpm2 Enpep Defb1 Ptprc Pecam1 Myh11 Nphs1 Umod Pi16 Pcp4l1 Synpo Wfdc15b S100a6 Sfrp2 Npnt Slc12a3 Pbx1 Rergl Wt1 Sfrp1 Pln Pard3b Ldhb Ren1 Ptpro Hsd11b2 Iqgap2 Kcnj1 Mafb Supplemental Table 3. Top variable genes per cluster

Full dataset Endothelium.avg_logFC Endothelium.gene Immune.avg_logFC Immune.gene Mesangium.avg_logFC Mesangium.gene Podocytes.avg_logFC Podocytes.gene Tubules.avg_logFC Tubules.gene 3.18 Ly6c1 4.1 Cd74 3.75 Acta2 2.53 Nphs2 4.02 Fxyd2 3.07 Srgn 4.1 H2-Aa 3.43 Myl9 2.48 Cdkn1c 3.96 Klk1 2.87 Emcn 3.9 H2-Ab1 3.40 Rgs5 2.16 Clic3 3.89 Defb1 2.77 Egfl7 3.5 H2-Eb1 3.38 Tagln 2.14 Nupr1 3.85 Atp1b1 2.74 Cdkn1a 3.0 Lyz2 3.09 Tpm2 1.99 Dpp4 3.82 2.71 Ifitm3 2.6 Tyrobp 2.96 Myh11 1.98 Enpep 3.45 Pvalb 2.70 Ednrb 2.5 Lilrb4a 2.93 Sparcl1 1.93 Tcf21 3.28 Spp1 2.60 Apold1 2.5 Laptm5 2.73 Mustn1 1.91 Nphs1 3.24 Umod 2.58 Igfbp5 2.4 Ccl9 2.72 Gm13889 1.87 Gadd45a 3.23 Slc12a3 2.57 Ctla2a 2.3 Fcer1g 2.53 Serpine2 1.86 Rab3b 3.14 Mt1 2.49 Cd24a 2.2 Cd52 2.45 Hopx 1.76 Rhpn1 3.10 Ldhb 2.39 Emp1 2.0 Ctss 2.38 Akr1b7 1.75 Tmsb4x 3.08 Wfdc15b 2.31 Ly6e 1.8 Ptprc 2.27 Pcp4l1 1.75 Col4a3 3.06 Sfrp1 2.24 S100a6 1.8 Cd53 2.18 Map3k7cl 1.74 Rasl11a 3.05 Calb1 2.23 Adgrf5 1.9 Cd83 2.10 Rergl 1.74 Mafb 2.87 Mt2 2.19 Flt1 2.3 Cytip 1.83 Pln 1.72 Npnt 2.70 Nudt4 2.16 Ehd3 1.4 Lsp1 1.82 Fxyd1 1.70 Arhgap24 2.66 Tmem213 2.12 Clic4 2.0 Srgn 2.16 Rasd1 1.70 Adm 2.55 Ppp1r1a 2.11 Crip1 1.9 Lgals3 2.15 Mgp 1.69 Pak1 2.55 Clu 2.06 1810011O10Rik 2.1 Rgs2 1.88 Sncg 1.68 Synpo 2.50 Tmem52b 2.05 Gimap6 1.7 Gm2a 1.67 Mef2c 1.65 Foxd2os 2.36 Kng2 2.04 Plpp1 1.5 Ly6e 1.64 Flna 1.63 Golim4 2.34 Egf 2.02 Slfn5 1.3 Fxyd5 2.13 Id3 1.61 Igfbp7 2.29 Wfdc2 1.94 Fxyd5 1.5 Ctsz 1.86 Gng11 1.60 Vegfa 2.22 Atp1a1 1.94 Pecam1 1.4 Gm8995 2.49 Cald1 1.59 Cd59a 2.17 mt-Cytb 1.94 Gm8995 1.3 Rpl18a 1.87 Dkk2 1.59 Sdc4 2.15 Gabarapl1 1.94 Pi16 1.4 Cyba 1.86 Nr4a2 1.58 Sema3g 2.14 Fabp3 1.93 Gimap4 1.4 Arpc1b 1.87 Rgs2 1.57 Tdrd5 2.09 Kl 1.92 Sgk1 1.5 B2m 1.79 Slc12a2 1.56 Nap1l1 2.04 Mal 1.92 Kdr 1.4 Gpx1 1.88 Filip1l 1.54 Shisa3 2.04 Kcnj1 1.90 Ptprb 1.3 Gm15427 1.47 Usp2 1.54 Eif3m 2.01 Pgam2 1.89 Ramp2 1.3 Emp3 1.77 Crip1 1.48 Thsd7a 1.98 mt-Nd1 1.87 1.4 Rpl32-ps 1.43 Akap12 1.48 Pth1r 1.97 mt-Co1 1.86 Pbx1 1.4 H2afz 1.52 Id2 1.48 Sept11 1.90 S100a1 1.86 Thbd 1.3 H2-K1 1.50 Mt1 1.47 Ctsl 1.90 Oxct1 1.85 Cyp4b1 1.3 Rpsa 1.07 Cdkn1a 1.46 Podxl 1.89 Chchd10 1.84 B2m 1.3 Rps4x 1.78 Impdh2 1.45 Cryab 1.87 Wnk1 1.83 Calcrl 1.2 Gm10689 1.32 Adamts1 1.44 Mertk 1.87 mt-Nd4 1.78 Ly6a 1.3 Gm29228 1.39 Igfbp5 1.44 Htra1 1.84 Slc25a5 1.78 Meis2 1.6 Rilpl2 1.50 Ppp1r12a 1.42 Nes 1.84 Sostdc1 1.78 Hbegf 1.2 Rps13-ps2 1.20 Tpm1 1.52 Wt1 1.83 mt-Nd2 1.76 Lrrc32 1.2 Gm9354 1.38 Actn1 1.55 Npr3 1.83 Epcam 1.76 Crip2 1.6 Msn 1.41 Emd 1.49 Ildr2 1.80 Clcnkb 1.70 Sdpr 1.3 Rpl23 1.19 Mylk 1.52 Robo2 1.80 Slc16a7 1.63 Clic1 1.2 Rpl32 1.08 Slc25a4 1.54 Pard3b 1.78 mt-Nd5 1.54 Nrp1 1.3 Gm9794 1.10 Gadd45b 1.48 Tmem150c 1.77 mt-Rnr1 1.48 Gpx1 1.2 Gm6472 1.13 Dstn 1.54 Gas1 1.77 Acat1 1.51 Plpp3 1.3 Rap1b 1.14 Cebpb 1.46 Hoxc8 1.76 Gm5514 1.40 Akap13 1.2 Rps24 1.07 Map1lc3a 1.50 Iqgap2 1.75 Pdzk1ip1 1.37 Mat2a 1.3 Cst3 1.13 Cstb 1.44 Sema3e 1.74 Ly6a Supplemental Table 3. Top variable genes per cluster

Endothelium subclusters residual doublets 0.avg_logFC 0.gene 1.avg_logFC 1.gene 2.avg_logFC 2.gene 3.avg_logFC 3.gene 4.avg_logFC 4.gene 0.85 Plat 2.09 Calca 0.98 Klf2 3.10 Mgp 1.96 Tcf21 0.83 Ednrb 1.98 Tspan8 1.20 Hspa1a 2.37 Fbln5 2.26 Clic3 0.96 Ctla2a 1.61 S100a4 1.23 Hspa1b 2.22 Jag1 2.63 Nphs2 0.71 Egfl7 0.83 Crip1 0.85 Jun 1.81 Igf2 1.78 Gsta4 0.51 Cd24a 0.89 Tm4sf1 0.99 Hsp90aa1 1.82 Cldn5 1.79 Ptpro 1.12 Akap12 0.99 Vim 1.34 Hsph1 1.87 Cxcl1 1.72 Rab3b 0.49 Igfbp5 1.24 Sox17 0.84 Fos 1.50 Egr1 1.78 Myom2 0.64 Marcksl1 0.68 S100a10 0.96 Sgk1 1.60 Fbln2 2.21 Gas1 0.50 Mat2a 1.08 Plac8 0.78 Egr1 1.31 Vim 2.14 Enpep 0.59 Litaf 0.76 S100a6 0.88 Ehd3 1.26 Klf2 1.87 Dpp4 0.58 Kdr 0.63 Ly6a 0.70 Fosb 1.28 Jun 1.53 Pdpn 0.31 Emcn 0.91 Palmd 1.02 Dnajb1 1.30 Fos 2.47 Igfbp7 0.61 S1pr1 0.77 Thbd 0.95 Ppp1r15a 0.83 Tmsb4x 1.53 Foxd2os 0.39 Ehd3 0.69 Slc6a6 0.93 Id3 1.18 Cd93 1.83 Gadd45a 0.58 Cyp4b1 0.75 Glul 0.69 Junb 1.15 Zfp36 2.53 Cdkn1c 0.37 Cdkn1a 0.66 Calcrl 0.91 Gm12346 1.19 Ier2 2.11 Sdc4 0.42 Tpm3 0.56 Fxyd5 0.92 Socs3 1.24 Id3 1.67 Tdrd5 0.45 Uqcrq 0.53 Anxa2 0.85 Zfp36 1.21 Stmn2 1.61 Pth1r 0.55 Lrrc32 0.39 Ifitm3 0.53 Hsp90ab1 1.21 Cyr61 1.84 Arhgap24 0.30 Lmna 0.57 Ahnak 0.98 Rhob 1.22 Slco2a1 1.73 Shisa3 0.33 Plpp1 0.52 Arpc1b 1.12 Plk2 0.84 Ly6a 1.56 Wt1 0.58 Cdk11b 0.60 Tinagl1 0.70 Hspa8 1.30 Fam3c 1.55 Itgb5 0.26 Srgn 0.50 Akap13 0.47 Klf4 1.04 Aggf1 1.65 Nphs1 0.42 Plpp3 0.53 S100a16 0.60 Igfbp5 0.99 Id2 1.41 Rhpn1 0.43 Tpm3-rs7 0.51 Uaca 0.50 Atf3 0.96 Utrn 2.12 Nupr1 0.32 Nrp1 0.39 Syn3 0.56 Dnaja1 1.06 Ptrf 1.47 Cryab 0.46 Hes1 0.45 Adgrl4 0.52 Hspb1 1.01 Sema3g 1.54 Mafb 0.31 Hbegf 0.35 Anxa1 0.69 Btg2 0.95 Nr4a1 1.55 Ctgf 0.35 Ppp1r2 0.44 Ifi203 0.69 Sertad1 0.76 Atf3 1.51 Ezr 0.30 Srsf2 0.40 1810011O10Rik 0.47 Plpp1 0.56 Ly6c1 1.39 Gm7658 0.37 Zcchc6 0.52 Cltb 0.58 Kdr 0.99 Tppp3 1.91 Rasl11a 0.27 Eif5 0.30 Gimap6 0.56 Plpp3 0.80 Ptprb 1.64 Cd59a 0.30 Flt1 0.42 Tspo 0.68 Dlc1 0.88 Gja4 1.36 Col4a4 0.28 Serinc3 0.44 Prr13 0.46 Ier2 0.71 Ppp1r15a 1.40 Pak1 0.37 Arpc3 0.31 Crip2 0.65 Id1 0.60 Fosb 1.62 Eif3m 0.34 Fkbp1a 0.28 Emp1 0.46 Nrp1 0.73 Cst3 1.68 Adm 0.29 Tax1bp1 0.31 Gpx1 0.54 Hsp25-ps1 0.88 Id1 1.38 Synpo 0.34 Sfpq 0.42 Atp5j 0.43 Dusp1 0.59 S100a4 1.62 Itm2b 0.29 Ywhab 0.32 Nupr1 0.56 Hbegf 0.80 Hexim1 1.52 Tmsb4x 0.26 Gabarap 0.29 Pi16 0.63 Adamts1 0.66 H2-K1 1.43 Tmsb10 0.35 Top1 0.28 Rpl11 0.51 Gng11 1.26 Fabp4 1.44 Palld 0.33 Map7d1 0.28 S100a11 0.51 2410006H16Rik 0.70 Rhob 1.54 Vegfa 0.38 Kitl 0.34 Cd200 0.48 Grasp 0.70 Heg1 1.34 Hist1h2bc NA NA 0.29 Klf3 0.43 Ece1 0.71 Sat1 1.43 Golim4 NA NA 0.30 Rpl30 0.44 Gm6368 0.55 Aplp2 1.37 Mpp5 NA NA 0.27 Rpl7 0.50 Nr4a1 0.82 Ehd4 1.40 Tspan13 NA NA 0.34 Rab11a 0.48 Hexim1 0.57 Ier3 1.43 Thsd7a NA NA 0.28 Adgrf5 0.42 Pcna 0.60 Btg2 1.35 Robo2 NA NA 0.27 Ly6e 0.46 Tek 0.70 Gadd45b 1.35 Lpin2 NA NA 0.28 Gm10689 0.36 Ncl 0.72 Gadd45g 1.47 Fnbp1l Supplemental Table 3. Top variable genes per cluster

Podocyte subclusters 0.avg_logFC 0.gene 1.avg_logFC 1.gene 2.avg_logFC 2.gene 3.avg_logFC 3.gene 4.avg_logFC 4.gene 5.avg_logFC 5.gene 0.66 Gm37376 0.50 Tnfrsf12a 0.42 Col4a3 3.37 Gm23935 1.84 Hbegf 2.40 Gm10800 0.77 Gm26669 0.36 Nupr1 0.26 Jun 2.50 Lars2 1.83 Lsp1 1.80 Gm10801 0.61 Syn3 0.39 Ifrd1 0.28 Podxl 2.01 Gm15564 1.71 Pla2g7 0.37 Gm37376 0.40 Gm42418 0.44 Anxa1 0.33 Hspa5 1.11 Neat1 1.30 Adamts1 0.43 mt-Nd6 0.36 Ifrd1 0.30 Nfkbia 0.25 Plat 0.42 mt-Rnr2 0.95 Cald1 0.33 Zcchc6 0.36 Klf6 0.30 Gm42418 0.29 Dpp4 0.64 Podxl 0.74 Ctgf 0.26 Bclaf1 0.51 Zfp36l1 0.26 Rps27 0.26 Hsp90b1 0.70 Sema3g 0.51 Tsc22d1 0.28 Gm26669 0.30 Nfkbia 0.33 Ier3 0.27 Enpep 0.54 Enpep 0.64 Cyr61 NA NA 0.39 Lmna 0.29 Rpl37 0.25 Sema3g 0.59 Nphs1 0.50 B2m NA NA 0.47 Gas1 0.35 Rasl11a 0.30 Thsd7a 0.49 mt-Rnr1 0.68 Slc6a6 NA NA 0.55 Foxp1 0.37 Gm26532 0.25 Calr 0.55 Npnt 0.33 Atf3 NA NA 0.51 Rn18s-rs5 0.32 Gadd45b 0.31 Pdia4 0.51 Dpp4 0.26 Vim NA NA 0.63 Kcnq1ot1 0.29 Rpl22l1 0.25 Htra1 0.51 mt-Nd5 0.36 Cystm1 NA NA 0.49 Arl13b 0.25 Rpl41 0.27 Aplp2 0.45 Plat 0.47 Cryab NA NA 0.27 Btg2 0.28 Hspe1 0.28 Sema3e 0.40 Vegfa 0.39 Tagln2 NA NA 0.41 Arid5b 0.26 Adm 0.27 Ctnna1 0.46 Col4a3 0.41 Chpt1 NA NA 0.31 Btg1 0.26 Dnaja1 0.26 Myo1e 0.54 Col4a4 0.29 Gadd45b NA NA 0.39 Nktr 0.31 Cdk2ap2 0.25 Scd2 0.53 Thsd7a 0.34 Sec62 NA NA 0.35 Myadm 0.28 Hspb1 NA NA 0.36 Sparc 0.35 Rpl18a NA NA 0.35 BC005537 0.35 Crip1 NA NA 0.57 Ptpro 0.26 H3f3b NA NA 0.33 Arl4a 0.28 Maff NA NA 0.35 mt-Cytb 0.37 H2-K1 NA NA 0.32 Rb1cc1 0.26 Ddit3 NA NA 0.53 Dag1 0.29 Gabarapl1 NA NA 0.26 Luc7l3 NA NA NA NA 0.39 mt-Nd1 0.36 Mtdh NA NA 0.33 Rock1 NA NA NA NA 0.43 mt-Nd2 0.25 Ppp1r15a NA NA 0.29 Erdr1 NA NA NA NA 0.37 mt-Nd4 0.30 Rpl38 NA NA 0.27 Zbtb20 NA NA NA NA 0.52 Alcam 0.40 Tuba1a NA NA 0.27 Rock2 NA NA NA NA 0.37 Itgb1 0.26 H2-D1 NA NA NA NA NA NA NA NA 0.48 Sema3e 0.30 Arid5b NA NA NA NA NA NA NA NA 0.44 Npr3 0.30 Gm26669 NA NA NA NA NA NA NA NA 0.49 Efnb1 0.29 Csrp2 NA NA NA NA NA NA NA NA 0.52 Angptl2 0.28 Loxl2 NA NA NA NA NA NA NA NA 0.44 Itgb5 0.27 Litaf NA NA NA NA NA NA NA NA 0.47 Pros1 0.28 Tspan3 NA NA NA NA NA NA NA NA 0.47 Ddn 0.34 Hnrnph1 NA NA NA NA NA NA NA NA 0.45 C1qtnf1 NA NA NA NA NA NA NA NA NA NA 0.42 Plod2 NA NA NA NA NA NA NA NA NA NA 0.40 Pdpn NA NA NA NA NA NA NA NA NA NA 0.40 P3h2 NA NA NA NA NA NA NA NA NA NA 0.39 Cldn5 NA NA NA NA NA NA NA NA NA NA 0.48 Aebp1 NA NA NA NA NA NA NA NA NA NA 0.47 Scd2 NA NA NA NA NA NA NA NA NA NA 0.40 Magi2 NA NA NA NA NA NA NA NA NA NA 0.43 Bcam NA NA NA NA NA NA NA NA NA NA 0.36 Aplp1 NA NA NA NA NA NA NA NA NA NA 0.35 Pdia6 NA NA NA NA NA NA NA NA NA NA 0.47 Ptprd NA NA NA NA NA NA NA NA NA NA 0.36 Dst NA NA NA NA NA NA NA NA NA NA 0.35 F2r NA NA NA NA NA NA NA NA NA NA 0.41 Tmbim6 NA NA NA NA NA NA NA NA NA NA 0.45 Ifngr1 NA NA NA NA Supplemental Figure 2

Endothelium (known) Podocytes (known) Supplemental Figure 3

Endothelium subclusters colored by replicate Subcluster composition A B 0 1

doublets

Replicate 2 3

4

Replicate

C Subcluster 1Subcluster Subcluster 2 Subcluster Subcluster 0 Subcluster

D S100a4 Thbd Hspa1b Fbln2 Supplemental Figure 3 continued

E

gene cluster 54 "cell maturation" enriched in cluster 2 and 0

gene cluster 34 "stress response" enriched in cluster 3 and 2

gene cluster 20 "cell adhesion" enriched in cluster 3 and 1

gene cluster 52 "endothelial cell proliferation" enriched in cluster 3 and 2 Supplemental Table 4. (Related to Fig. 3D and Suppl. Fig. 3E) Gene sets underlying the gene ontology (GO) term analysis for glomerular endothelium subclusters Barplots show the top 20 term which passed a p-value cutoff of 0.01

geneCluster.54 enriched in subcluster 2 and 0geneCluster.34 enriched in subcluster 3 and 2geneCluster.20 enriched in subcluster 3 and 1geneCluster.52 enriched in subcluster 3 1 Akap12 2410006H16Rik 1700025G04Rik Aggf1 2 App Atf3 Adam15 Arhgap44 3 Arap2 Btg2 Atox1 Arl15 4 Arf4 Cebpd Calca Art3 5 Arrdc3 Dnaja1 Calcrl AW112010 6 Atp1a1 Dnajb1 Cd93 Bmp2 7 BC005537 Dusp1 Crip1 Casp1 8 Cd14 Egr1 Cyb5r3 Cbr2 9 Cd24a Fos Edn1 Cdh13 10 Cd300lg Fosb Ehd4 Cgnl1 11 Cdk11b Gadd45b Emp3 Cldn5 12 Cdkn1a Gas5 Fam107a Cox20 13 Crem Gata2 Fkbp3 Ctsh 14 Csrnp1 Gm12346 Fxyd6 Cxcl1 15 Ctla2a Gm15542 Gja4 Cxcl2 16 Ctsl Gm6368 Ifi27l2a Dhh 17 Cyp4b1 Hexim1 Itgb1 Dst 18 Ddx3y Hsp25-ps1 Ly6a Egfl8 19 Dnajb9 Hsp90aa1 Ly6e Fbln2 20 Ednrb Hsp90ab1 Mprip Fbln5 21 Egfl7 Hspa1a Palmd Flrt1 22 Ehd3 Hspa1b Plac8 Ganc 23 Eif1 Hspa5 Plec Gbp3 24 Eif1-ps1 Hspa8 Polr2e Gm4759 25 Eif4a-ps4 Hspb1 Ptrf H2-Ab1 26 Eif4a1 Hsph1 S100a4 Htra1 27 Emcn Icam1 Serpinb1a Igf2 28 Ets2 Id1 Slc6a6 Igfbp4 29 Fam167b Id2 Ssu2 Jag1 30 Gm10157 Id3 Thbd Kcne3 31 Gm20568 Ier2 Tm4sf1 Ltbp4 32 Gmfb Ier3 Tmsb4x Mgp 33 Gngt2 Irf1 Tppp3 Mgst1 34 Hbegf Jun Tspan8 Npnt 35 Hes1 Junb Tspo Nuak1 36 Ifrd1 Klf2 Vim Pir 37 Igfbp5 Klf4 Prss23 38 Ikbkb Nfkbia Pthlh 39 Ipmk Nr4a1 Rsad2 40 Irf8 Phlda1 Stat5a 41 Isy1 Plk2 Stc1 42 Kdr Ppp1r15a Stmn2 43 Litaf Rhob Top2b 44 Lmcd1 Sertad1 Vcam1 45 Marcksl1 Sgk1 Vegfc 46 Mat2a Socs3 Wfdc1 47 Midn Tob1 Wwp1 48 mt-Cytb Tsc22d1 49 mt-Nd1 Ubc 50 Mt1 Zfas1 51 Myof Zfp36 52 Nhp2l1 53 Nktr 54 Nostrin 55 Nr4a2 56 Nrp1 57 Ntn4 58 Paip2 59 Pbx1 60 Plat 61 Plaur 62 Plpp1 63 Plpp3 64 Prkab1 65 Ramp3 66 Rell1 67 Sec62 68 Slc43a3 69 Slfn5 70 Smarca5 71 Spry4 72 Stat3 73 Timeless 74 Tiparp 75 Tpm3 76 Tpm3-rs7 77 Tra2a 78 Uqcrq Supplemental Figure 4

A 5 B

0

2

1

4 3 C

6 Supplemental Figure 5

A Podocyte subclusters colored by replicate B Subcluster composition

0 1

Repl. 2 3

4 5

Replicate

C D Lars2 non-podoc. podocytes nuclei Subcluster 3 Subcluster

IgG control non-podoc. podocytes nuclei Subcluster 5 Subcluster Supplemental Figure 6

A

B

gene cluster 38 gene cluster 55

Aldh1a1 Pdlim4 Gm15564 Bmp4 Pla2g7 Gm23935 Cd200 Plppr4 Lars2 Csflr Prss23 Dkk3 Ptgis Efemp1 Rasl11b Fam180a Rbm1 Hey2 Smim5 Igfbp2 Tgm2 Lsp1 Tmem176b Mustn1 Wwc1 Ntn4 Supplemental Figure and Table legends (order as they appear in the main text)

Supplemental Figure 1. (Related to Fig. 1) Quality of glomeruli single-cell data and comparisons with bulk mRNAseq data (A) Representative brightfield microscopic images of isolated mouse glomeruli showing high purity of glomeruli. Scale bar corresponds to 100 µm. Magnified areas indicate a tubular fragment (left), a decapsulated glomerulus with protruding vasculature (middle) and an avascular, decapsulated glomerulus (right). Quantification of 123 structures in a 10µl droplet of a glomerular preparation: tubulular fragments 13% (n=16), vascular glomeruli 13% (n=16), avascular glomeruli 74% (n=91), Bowman’s capsules 0% (n=0) (B) Distribution and median of the number of genes and UMIs detected per cell in four separate Drop-seq runs corresponding to independent biological replicates (left panel) or to a pool of replicates (right panel). Cells were filtered to express >250 genes and >350 UMIs. In the pool, cell doublets were removed computationally (see Methods and Suppl. Table 1 for median number of genes, UMIs and reads per cell). (C) tSNE plot from Fig. 1B with each cell now coloured according to biological replicate. The batch effect visible in the podocyte cluster was regressed out for subclustering analysis (Fig. 4; Suppl. Fig. 5A, B). (D) Heat map indicating that samples cluster by cell type, not by experimental replicate. The fraction of each cell type per replicate is indicated in the bar plot on top. (E) Pairwise correlations between Drop-seq single-cell data and bulk mRNA seq samples of similar depth generated in this study. "bulk1" corresponds to glomerular lysate (before single- cell dissociation) and "bulk2" corresponds to single-cell suspensions prepared by enzymatic digestion followed by live cell sorting by flow cytometry. The analysis showed high correlations in general, but also indicated that single-cell data correlate slightly better with the bulk data from cells that underwent single-cell dissociation (circles). (F) Scatterplots showing the expression in the "bulk1" and "bulk 2" samples for the 50 most variable genes per cell type identified in the single-cell RNAseq data. The data suggest that podocytes were overrepresented in the "bulk2" sample, indicating that cell dissociation introduced a bias. Endothelial and mesangial cells showed a higher dispersion with the opposite tendency, while tubules were not affected. (G) Boxplots of expression ratios in the bulk data for the 50 most variable genes per cell type (corresponding to (F)). A distribution along the horizontal line would indicate absence of bias. (H) Pairwise correlations between cell types identified by single-cell data generated in this study (pre-fix "ds") and published bulk mRNA-seq data sets. The analysis showed the highest correlations with the corresponding cell type, supporting cell type assignments in the single-cell data. Scale for the heatmap is indicated. R = Pearson correlation efficient. mRNAseq data were from Brunskill et al. 201412: gsm1585030 (proximal tubules), gsm1585043 (MafB podocytes), gsm1585044 (mesangial cells), gsm1585047 (endothelial cells); from Kann et al.1: gsm156360 (podocytes).

Supplemental Table 1. (Related to Suppl. Fig. 1B). Median number of genes, transcripts (UMIs) and reads per cell This table lists the number of uniquely mapped reads per cell, the number of cells, the number of detected genes and the number of transcripts per cell (defined as unique molecular identifiers (UMIs). Numbers are given for data before or after cell doublet removal.

Supplemental Table 2. (Related to Suppl. Fig. 1B). Marker genes for each of the five major cell types to identify putative cell doublets Cell-type specific genes as listed in this table were used to identify putative cell doublets (see Online Methods for details).

Supplemental Table 3. (Related to Suppl. Fig. 1B, Fig. 3A and Fig. 4A). Top variable genes per cell cluster or subcluster This table consists of three subtables which list the 50 most highly variable genes in each cell cluster as determined by the function 'FindAllMarkers' in Seurat6. 'Full dataset' corresponds to Fig. 1B (cell type clusters). 'Endothelium' corresponds to Fig. 3A (endothelium subclusters). 'Podocytes' corresponds to Fig. 4A (podocyte subclusters).

Supplemental Figure 2. (Related to Fig. 2) Additional markers of glomerular cells Distribution and relative expression of individual highly variable genes (violin plots) in endothelium and podocytes. Established markers (bold) and markers described in the literature (italics).

Supplemental Figure 3. (Related to Fig. 3) Subclustering of glomerular endothelium (A) 2D representation corresponding to the plot in Fig. 3A. Replicate Drop-seq experiments 1 to 4 contributed to all clusters. (B) Waffle plots show that each cluster contained cells from all four replicates, while the composition of cluster 4, identified as residual cell doublets, was biased. A square represents 1% of the total number of cells in a cluster. (C) Expression of markers colored based on normalized expression levels. Upper panels correspond to subcluster tSNE plot shown in (A), lower panels to the tSNE plot of the whole data set as shown in Fig. 1B. (D) Immunohistochemistry images obtained from Human Protein Atlas (HPA)10 support that endothelial subcluster marker proteins S100a4, Thbd, Hspa1b, and Fbln2 may be expressed heterogeneously in human glomerular endothelium. Area shown (500 x 500 pixels = 200 µm2) corresponds to glomeruli taken from a larger HPA image. Red arrows point to endothelial cells as identified by intracapillary nuclei with strong (S100a4) or positive (Thbd, Hspa1, Fbln2) marker expression. Green arrows point to endothelial cells with weak (S100a4) or no (Thbd, Hspa1, Fbln2) marker gene expression. Magnified areas (50 x 50 pixels = 20 µm2) are shown below with red squares representing strong/positive expression and green squares weak or no expression. (E) Gene ontology (GO) terms corresponding to heat map shown in Fig. 3C. The heatmap indicates that 4 endothelial subclusters (0: red; 1: green; 2: blue; 3: violet) show distinct, overrepresented gene activation patterns (see also Suppl. Fig. S3C): Cluster 1 and 3 (gene set 20 terms related to “cell adhesion”), Cluster 2 and 0 (gene set 54 terms related to “cell maturation”), cluster 2 and 3 (gene set 34 terms related to ”stress response”), and cluster 3 (gene set 52 terms related to “endothelial cell proliferation”; see panel E). p values were calculated by hypergeometric testing (Fisher's exact test). Corresponding gene clusters are listed in Suppl. Table 4.

Supplemental Table 4. (Related to Suppl. Fig. 3D and Suppl. Fig. 3E). Gene sets underlying the gene ontology (GO) term analysis for glomerular endothelium subclusters This table lists the sets of genes, expressed in endothelium subclusters, that underlie the GO analysis shown in Fig. 3D and Suppl. Fig. 3E.

Supplemental Figure 4. (Related to Fig. 4) Stress response in podocytes (A) Upper panel: 2D tSNE representation of podocyte subclustering before correcting for tissue-dissociation-induced stress response genes7. Subcluster 4 showed upregulation of stress response genes. Lower panel: Expression of stress response markers colored based on normalized expression levels (grey: low, red: high). (B) Stress response genes, known to become transcriptionally induced by tissue- dissociation7, show higher expression in bulk mRNA-seq samples from glomerular cells that underwent prolonged enzymatic tissue dissociation ("bulk2") as compared to non-dissociated glomeruli ("bulk1"). The set of 140 genes was taken from van den Brink et al. 20177, Suppl. Table S5: Actg1, Ankrd1, Arid5a, Atf3, Atf4, Bag3, Bhlhe40, Brd2, Btg1, Btg2, Ccnl1, Ccrn4l, Cebpb, Cebpd, Cebpg, Csrnp1, Cyr61, Cxcl1, Dcn, Ddx3x, Ddx5, Des, Dnaja1, Dnajb1, Dnajb4, Dusp1, Dusp8, Erf, Egr1, Egr2, Eif1, Eif5, Errfi1, Fam132b, Fos, Fosb, Fosl2, Gadd45a, Gadd45g, Gcc1, Gem, H3f3b, Hipk3, Hspa1a, Hspa1b, Hspa5, Hspa8, Hspb1, Hspe1, Hsph1, Hsp90aa1, Hsp90ab1, Idi1, Id3, Ier2, Ier3, Ier5, Il6, Irf1, Ifrd1, Itpkc, Jun, Junb, Jund, Kcne4, Klf2, Klf4, Klf6, Litaf, Lmna, Maff, Mafk, Mcl1, Midn, Mir22hg, Mt1, Mt2, Myadm, Myd88, , Nckap5l, Ncoa7, Nfkbia, Nop58, Nppc, Nr4a1, Odc1, Osgin1, Oxnad1, Pcf11, Per1, Phlda1, Pnrc1, Pnp, Ppp1cc, Ppp1r15a, Pxdc1, Rassf1, Rhob, Rhoh, Ripk1, Sat1, Sbno2, Sdc4, Skil, Slc10a6, Slc38a2, Slc41a1, Socs3, Sqstm1, Srf, Srsf7, Stat3, Tagln2, Tiparp, Tnfaip3, Tnfaip6, Tpm3, Tra2a, Tra2b, Trib1, Tubb4b, Tubb6, Usp2, Irf8, Klf9, Nfkbiz, Pde4b, Rap1b, Serpine1, Srsf5, Tppp3, Ubc, Wac, Zc3h12a, Zfand5, Zfp36, Zfp36l1, Zfp36l2, Zyx. (C) Quantitation of the tissue-dissociation-induced stress response between "bulk2" and "bulk1" mRNA-seq samples for genes with a normalized expression greater than 1 (132 out of 140). Stress-response genes showed significantly elevated expression in single, dissociated cells compared to comparable sets of random genes or highly variable genes (for details see Methods; t-test p<2.2e-16 for each comparison).

Supplemental Figure 5. (Related to Fig. 4) Subclustering of podocytes (A) Upper panel: 2D representation of podocyte subclustering after correcting for stress response by regressing out stress response genes. Cells are colored by replicates (1 to 4). Replicates contribute to all clusters. Lower panel: Expression of stress response markers colored based on normalized expression levels (grey: low, red: high). (B) Waffle plots showing that each cluster contained cells from all four replicates. A square represents 1% of the total number of cells in a cluster. (C) Expression of markers colored based on normalized expression levels. Upper panels correspond to subcluster tSNE plot as shown in (Fig. 3A), lower panels to the tSNE plot of the whole data set as shown in Fig. 1B. (D) Laser scanning confocal microscopy of glomeruli purified from kidneys of transgenic Nphs2-Cre x mT/mG double-fluorescent Cre reporter mice11. Podocytes are genetically marked by mG (Cre-dependent membrane-targeted GFP; green) fluorescence, while non- podocyte cell types remain mT (membrane-targeted tandem dimer Tomato; red)-positive. Top row: purified, whole glomeruli. Lars2 indicates Lars2 antibody staining (left) or IgG control (right). Second row: Yellow arrowheads points to a Lars2-positive podocyte. Third row: Yellow arrowhead points to a Lars2-negative podocyte. Scale bars correspond to 10µm. Magnified areas are 22 x 22 µm2. IgG isotype control images showed no signal.

Supplemental Figure 6. (Related to Fig. 4) Pathway and gene set overdispersion analysis for podocyte subclusters (A) Pathway and gene set overdispersion analysis (PAGODA)8 of the podocyte subclusters shown in Fig. 4A,B. The heatmap shows distinct, overrepresented gene activation patterns only for subclusters 4 (blue) and 3 (turquoise). (B) Gene ontology (GO) terms corresponding to the heat map in (A). Podocyte subclusters 4 and 3 express gene sets that had no associated GO terms.

References for Supplemental Material

1. Kann M, Ettou S, Jung YL, Lenz MO, Taglienti ME, Park PJ, Schermer B, Benzing T, Kreidberg JA: Genome-Wide Analysis of Wilms’ Tumor 1-Controlled Gene Expression in Podocytes Reveals Key Regulatory Mechanisms. J. Am. Soc. Nephrol. JASN 26: 2097– 2104, 2015

2. Macosko EZ, Basu A, Satija R, Nemesh J, Shekhar K, Goldman M, Tirosh I, Bialas AR, Kamitaki N, Martersteck EM, Trombetta JJ, Weitz DA, Sanes JR, Shalek AK, Regev A, McCarroll SA: Highly Parallel Genome-wide Expression Profiling of Individual Cells Using Nanoliter Droplets. Cell 161: 1202–1214, 2015

3. Alles J, Karaiskos N, Praktiknjo SD, Grosswendt S, Wahle P, Ruffault P-L, Ayoub S, Schreyer L, Boltengagen A, Birchmeier C, Zinzen R, Kocks C, Rajewsky N: Cell fixation and preservation for droplet-based single-cell transcriptomics. BMC Biol. 15: 44, 2017

4. Karaiskos N, Wahle P, Alles J, Boltengagen A, Ayoub S, Kipar C, Kocks C, Rajewsky N, Zinzen RP: The Drosophila embryo at single-cell transcriptome resolution. Science 358: 194–199, 2017

5. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR: STAR: ultrafast universal RNA-seq aligner. Bioinforma. Oxf. Engl. 29: 15– 21, 2013

6. Satija R, Farrell JA, Gennert D, Schier AF, Regev A: Spatial reconstruction of single-cell gene expression data. Nat. Biotechnol. 33: 495–502, 2015

7. van den Brink SC, Sage F, Vértesy Á, Spanjaard B, Peterson-Maduro J, Baron CS, Robin C, van Oudenaarden A: Single-cell sequencing reveals dissociation-induced gene expression in tissue subpopulations. Nat. Methods 14: 935–936, 2017

8. Fan J, Salathia N, Liu R, Kaeser GE, Yung YC, Herman JL, Kaper F, Fan J-B, Zhang K, Chun J, Kharchenko PV: Characterizing transcriptional heterogeneity through pathway and gene set overdispersion analysis. Nat. Methods 13: 241–244, 2016

9. Yu G, Wang L-G, Han Y, He Q-Y: clusterProfiler: an R package for comparing biological themes among gene clusters. Omics J. Integr. Biol. 16: 284–287, 2012

10. Uhlén M, Fagerberg L, Hallström BM, Lindskog C, Oksvold P, Mardinoglu A, Sivertsson Å, Kampf C, Sjöstedt E, Asplund A, Olsson I, Edlund K, Lundberg E, Navani S, Szigyarto CA-K, Odeberg J, Djureinovic D, Takanen JO, Hober S, Alm T, Edqvist P-H, Berling H, Tegel H, Mulder J, Rockberg J, Nilsson P, Schwenk JM, Hamsten M, von Feilitzen K, Forsberg M, Persson L, Johansson F, Zwahlen M, von Heijne G, Nielsen J, Pontén F: Proteomics. Tissue-based map of the human proteome. Science 347: 1260419, 2015

11. Muzumdar MD, Tasic B, Miyamichi K, Li L, Luo L: A global double-fluorescent Cre reporter mouse. Genes. N. Y. N 2000 45: 593–605, 2007

12. Brunskill EW, Park J-S, Chung E, Chen F, Magella B, Potter SS: Single cell dissection of early kidney development: multilineage priming. Dev. Camb. Engl. 141: 3093–3101, 2014

SIGNIFICANCE STATEMENT

The glomerular filtration unit consists of three tightly intertwined cell types: the endothelium, the mesan- gium, and podocytes. Despite established physiologic cues acting on these cells within a single glomerulus, cellular heterogeneity in the healthy glomerulus remains poorly characterized. To address this problem, the au- thors performed large-scale transcriptional profiling of 13,000 mouse glomerular cells. They provide a com- prehensive atlas of gene expression for the known glomerular cell types and describe potential sub- populations for endothelium and podocytes. An in- teractive, freely accessible web tool allows for querying and visualizing the data. The study highlights the power of single-cell RNA sequencing to study gene expression in the kidney and sets the stage for future investigations of glomerular dysfunction in disease.